>>Does Angel go off and probe for PCI devices?  If so a master abort is to be
>>expected.  Otherwise yes, I would agree it's odd.
>>
>Ah yes, I see this is the case. I guess I assumed that this condition would
>get cleared and thus if it persisted it means it is recurring (which is a
>problem?). Maybe its no big deal so nobody bothers to clear it?

Yeah.  The master abort is harmless in itself and the 21285 doesn't force you 
to clear it.  Angel almost certainly doesn't care.

>What should NetBSD's 21285 values be? Comments anyone please. I'd really
>appreciate it.

The ones you quoted look pretty reasonable to me.  You just need to check 
that the values that get handed on to devices match what is in the BARs.

p.